Skip to content

feat: default MiniMax examples to M3.1 Flash - #438

Open
octo-patch wants to merge 5 commits into
algorithmicsuperintelligence:mainfrom
octo-patch:feature/add-minimax-provider
Open

octo-patch wants to merge 5 commits into
algorithmicsuperintelligence:mainfrom
octo-patch:feature/add-minimax-provider

Conversation

@octo-patch

@octo-patch octo-patch commented Mar 15, 2026 •

Copy link
Copy Markdown

Reason: Make the MiniMax example select MiniMax-M3.1-Flash by default while retaining MiniMax-M3 as an opt-in ensemble member.

Changes

  • Add the MiniMax configuration with Flash enabled for evolution and evaluation. M3 has zero weight until explicitly enabled; older model IDs are no longer advertised in the example.
  • Document both regional endpoints, current context windows, and Flash effort settings. Remove the unsupported per-iteration cost estimate and stale temperature guidance.
  • Exercise configuration loading, ensemble selection, and real SDK serialization with an offline HTTP transport. Cover regional URLs, default requests, all five Flash effort levels, explicit M3 selection, saved/custom model IDs, and preservation of image/video content parts.
  • Preserve existing configuration semantics and the generic API transport. Explicit model names are never rewritten.

Validation

  • Python 3.10.12, installed with pip install -e ".[dev]"; Black 25.12.0 and isort 7.0.0 match the repository's hook pins.
  • python -m unittest discover -s tests -p test_minimax_config.py -v: 4 passed, including controller initialization and SDK serialization against mocked HTTP transport.
  • python -m unittest discover -s tests -p 'test_*.py' -v: 691 tests run, OK (2 circle-packing tests skipped because optional SciPy is not installed).
  • black --workers 2 --check tests/test_minimax_config.py tests/test_valid_configs.py: passed.
  • isort --check-only tests/test_minimax_config.py tests/test_valid_configs.py: passed.
  • git diff --check and diff secret scan: passed.

Tests ran in a checkout-local environment without inherited provider or automation credentials. The separate optillm/model integration suite was not run.

References

No paid inference is used. At head 679f5cc8685feff2b5c27e37d612589bfaa5fc92, the upstream Tests and Security Scan workflows report action_required, and license/cla is pending because the CLA is not signed. These checks require separate resolution before merge; hosted CI has not passed.

@CLAassistant

CLAassistant commented Mar 15, 2026 •

Copy link
Copy Markdown

CLA assistant check
Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
0 out of 2 committers have signed the CLA.

❌ PR Bot
❌ octo-patch


PR Bot seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
You have signed the CLA already but the status is still pending? Let us recheck it.

MiniMax offers OpenAI-compatible API with models like MiniMax-M2.5 and
MiniMax-M2.5-highspeed (204K context window). This commit adds:

- MiniMax provider section in README (setup guide, cost estimation)
- Example config file (configs/minimax_config.yaml)
- Updated test env to support MiniMax API key validation

Co-Authored-By: octo-patch <octo-patch@users.noreply.github.com>
@octo-patch
octo-patch force-pushed the feature/add-minimax-provider branch from 5545f60 to 64dc033 Compare March 15, 2026 06:04
@octo-patch

Copy link
Copy Markdown
Author

I've updated the commit author to be properly linked to my GitHub account. The CLA check should pass now. @CLAassistant check

- Update MiniMax-M2.7 and MiniMax-M2.7-highspeed as default models
- Update config, README, and docs references
- Keep all previous models as alternatives
@octo-patch octo-patch changed the title Add MiniMax as a supported LLM provider feat: add MiniMax as a supported LLM provider (M2.7 default) Mar 18, 2026
- Add MiniMax-M3 to model list and set as default
- Keep MiniMax-M2.7 and MiniMax-M2.7-highspeed as alternatives
- Update README and configs/README.md with M3 references

Co-Authored-By: Octopus <liyuan851277048@icloud.com>
@octo-patch octo-patch changed the title feat: add MiniMax as a supported LLM provider (M2.7 default) feat: upgrade MiniMax default model to M3 Jun 2, 2026
@octo-patch octo-patch changed the title feat: upgrade MiniMax default model to M3 feat: default MiniMax examples to M3.1 Flash Oct 20, 2026

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants